300 likes | 466 Views
A novel approach to detector calibration parameter determination and detector monitoring. Eduardo Rodrigues On behalf of the LHCb VELO Group CHEP 2010, Taipei, Taiwan, 21 October 2010. OUTLINE: LHCb , vertex detector Data processing Detector parameters optimisation.
E N D
A novel approach todetector calibration parameter determinationand detector monitoring Eduardo Rodrigues On behalf of the LHCb VELO Group CHEP 2010, Taipei, Taiwan, 21 October 2010 • OUTLINE: • LHCb, vertex detector • Data processing • Detector parameters optimisation
The LHCbexperiment @ the LHC Forward spectrometer Acceptance: ~10 – 300 mrad Luminosity: 21032 cm-2 s-1 Nr of B’s / 2fb-1(nominal year): 1012 Detector: excellent tracking excellent PID Reconstruction: - muons: easy - hadronic tracks: fine - electrons: OK - p0’s: OK, though difficult - neutrinos: no Mission statement - Search for new physics probing the flavour structure of the SM - Study CP violation and rare decays with beauty & charm hadrons b b B flight path of the order 5-10mm p p b b σ(inclusive) ~ 80 mb σ(b bbar) ~ 0.5 mb All b-hadron species produced
Tracking system requirements Vertexing • Precise reconstruction and separation of primary and secondary vertices- identification of beauty & charm meson decays Tracking • Excellent pattern recognition • Precise determination of track parameters excellent momentum resolutiondp/p = 0.35% to 0.55% excellent impact parameter (IP) resolutions p p Beauty & charm mesons Excellent - mass resolution 15-40 MeV - propertime resolution ~50fs
21 April 2010: first reconstructed beauty particle B+ J/Y(mm) K+ Candidate
Precisetracking and vertexing (1/2) Crucial to LHCb physics programme Hi resolution for VELO sensors • Hit resolution versus strip pitch for 2 bins of the projected angle • Evaluated for R sensors using residuals of long tracks • Hit resolution depend to 1st order on: • projected angle • strip pitch
Precisetracking and vertexing (2/2) Primary vertex resolution in Z Crucial to LHCb physics programme • Best such resolutions at the LHC IP = Impact Parameter p- Direction of B Primary vertex p+
The LHCbVErtexLOcator Highest precision vertex detector at the LHC
y x z y x VELO – VErtexLOcator base cooling pile-up modules injection p LHC vacuum stable beams Modules (21+2) RF box p • 2 retractable detector halves:~8 mm from beam when closed, retracted by 30mm during injection • 21 stations per half with an R and a sensor • 2 extra pile-up stations per half- recognition of multiple interaction collisions at the trigger level • # read-out channels: ~ 200k • Non-zero suppressed data rate ~ 4 GB/s !
VELO – modules • Purpose : • Hold the sensors fixed wrt module support • Connect electrical readout to the sensors and routing of signals to DAQ system • Provide means of cooling to the sensors • Sensor-sensor positioning accuracy < 5mm r-side circuit R-sensor carbonfibre -sidecircuit thermal pyrolyticgraphite -sensor
VELO – sensors R sensor F sensor • Highly segmented; n+ on n • 2048 strips per sensor • Design operation at -7 degrees • Read out at 1 MHz 512 strips 512 strips 1385 outer strips 683 inner strips 512 strips 512 strips Data
VELO data processing Achieving the best performance
Data processingchain VELO sensor TELL1 = DAQ board Event banks • VELO sends analogue signal digitised on TELL1s • TELL1s: 4 FPGAs with firmware, which work on integers(Why FPGAs? Faster / better / more reliable / cheaper) • Mimicking the integer operations on FPGAs: need for a software emulation
VELO data output (i.e. rawevent) types Zero-suppressed banks Non-zero suppressed (NZS) banks Error banks main input to the reconstruction - clusters from hits - 1 bank per sensor Full data as read out by a sensor - main input to emulation Contains information on errors produced by the TELL1 boards Raw event types (main ones) • ZS data rate: 50 bytes x 1 MHz = 50 Mb / s
TELL1 acquisition boardsprocessing = not used for now RAW Set of algorithms on TELL1 board - Data synchronisation, buffering- TELL1s first digitise the analogue signalsfrom the sensors- Suppress noise and performclusterisation - Algorithms implemented in FPGAs on TELL1s Pedestalssubtraction Front-End chip cross-talk correction Cable cross-talk correction • Algorithms require ~ 106 configuration parameters: • Most important are- Pedestals- Clusterisation thresholds • Need to be determined and optimised ! How … ? Common mode subtraction Channel Reordering Common mode subtraction Clusterisation Clusters
Detector parameterdetermination & optimisation Standard approach : • FPGA algorithm parameters typically determined via standalone calculations or measurements • But how to achieve the best performance given- complex chain of algorithms- very large number of configuration/calibration parameters Novel approach : • Integrate determination of detector parameters in the data processing framework itself- Full integration in the LHCbsoftware framework • Use non-zero-suppressed data output by the TELL1s and emulate the data processing to tune its calibration parameters
TELL1 processing and emulation TELL1 = DAQ board TELL1emulation NZS bank ~106 configuration parameters ! TELL1 processing Emulated ZS bank= emulated clusters (emu. = RAW bit perfectness) ZS bank = clusters Emulation + NZS data:means of determining and optimising the TELL1 processing parameters
The Vetra software project Main characteristics of Vetra : • LHCbsoftware project for the VELO detector (also silicon tracker) • Mimics the whole processing sequence • Implements in C an emulation of the TELL1 algorithms run on the FPGAs (VHDL) • Can treat as input the same data format(s) as output by the TELL1s- NZS, ZS, etc. • Highly flexible and configurable, with Python • Unique framework for the determination and tuning of the various types of TELL1 parameters Other benefits : • Allows for self-consistency checks bit-perfectness of the TELL1 emulation • Provides monitoring of all processing steps • Provides at the same time a software framework and tools for TELL1 studies (e.g. new algorithms) and the online and offline monitoring of the VELO • Also used in test-beam and laboratory tests
Database TELL1 parameterdetermination– procedure TELL1 board Noise NZS banks PRESENT Calibration parameters XML file Vetra XML file NEW parameters Bit-perfectemulation Analyses
Pedestal correction and its monitoring Raw Pedestal corrected Pedestal Bank
Detector monitoring and data quality • Uses the Vetra framework and “tools”
Experiencefrom the 2010 run • We coped with the increase in LHC luminosity(LHCb runs under different conditions wrt ATLAS/CMS) • VELO calibration performed with the emulation suite - New parameters determined and uploaded to the TELL1s acquisition boards when/if necessary allows us to cope with extra increases in rate, via regular monitoring and determination of new parameters • Approach successfully applied to collision data taken in 2010 - Integration in the data processing framework was the choice to make • Regular monitoring of the detector done on a daily basis - Noise, pedestals stability, hit and cluster reconstruction performance, etc. • The Vetra software framework has also been adopted by the silicon tracker group for their tracking stations
So, doesitwork? Primary Vertex Secondary Vertex
The LHCb detector Ring Imaging Cherenkov Calorimeters 250/300 mrad Acceptance 10 mrad pp collision (side view) Muon System « Tracking » detectors
Hit resolutions for VELO sensors • Hit resolution versus strip pitch for 2 bins of the projected angle • Evaluated for R sensors using residuals of long tracks • Hit resolution depend to 1st order on: • projected angle • strip pitch